US ‘bans’ Fable 5, Mythos 5; Anthropic issues statement

What the Directive Requires

Anthropic confirmed it received the directive on Friday at 5:21 p.m. ET. It mandates the immediate disabling of both Fable 5 and Mythos 5 for all foreign nationals, whether located inside or outside the United States. This includes Anthropic’s own foreign national employees.

The company quickly moved to comply with the legal order, ensuring access to its other AI models remains unaffected.

The Alleged Jailbreak

The government’s concern reportedly stems from a method of bypassing, or jailbreaking, Fable 5. Anthropic had recently released Fable 5 as a specialized version of its tightly controlled Mythos AI system, aiming to broaden access to advanced capabilities.

According to Anthropic, the directive did not provide specific details regarding the national security implications of this bypass. The company reviewed a demonstration of the alleged technique and found it identified minor, previously known software vulnerabilities.

Anthropic Pushes Back

In a comprehensive statement, Anthropic articulated its strong opposition to the government’s action. The company argues the identified jailbreak is narrow and does not demonstrate capabilities beyond those already available in other leading AI systems, including OpenAI’s GPT 5.5.

Anthropic highlighted its extensive efforts in red teaming Fable’s safeguards with the US government, the UK AISI, and various third party organizations for thousands of hours prior to launch. These tests, the company claims, showed Fable’s safeguards to be substantially more effective than previous models, with no universal jailbreak discovered to date.

The ‘Defense in Depth’ Strategy

Anthropic further explained that perfect jailbreak resistance is likely impossible for any model provider currently. The company adopted what it calls a defense in depth strategy for Fable 5, aiming to make jailbreaks either narrow or costly to produce, combined with thorough monitoring.

This strategy also led to a 30 day customer data retention policy, which Anthropic says carries real costs but aids in researching and mitigating potential bypasses. The company believes the government’s action sets an unfeasible standard that could halt new model deployments across the frontier AI industry. More information is expected within the next 24 hours.

What This Means for AI Governance

This incident underscores the escalating tension between rapid AI development and national security concerns, potentially setting a precedent for future regulatory interventions.

Anthropic expressed its belief that the government should have the ability to block unsafe deployments, but only through a statutory process that is transparent, fair, clear, and grounded in technical facts. The company has apologized to its customers for the disruption and is hoping to restore access as soon as the matter can be resolved.